This reference uses the specific documents linked in the article and listed below. Check their scope, publication dates and the statements they support.
Write the question before opening the dashboard
Decide whether the review concerns a specific service cohort, a farrowing cohort or events recorded during a calendar period. Those selections can contain different animals. A sow counted in one month’s service records need not contribute to that month’s farrowing records. Keep the date that selects a row separate from the date a worker entered it.
Name the measure in full beside its formula. A total per recorded litter and a total per sow enrolled answer different questions. Preserve the farm’s categories for liveborn, stillborn, other recorded outcomes and uncertain entries; do not invent a biological classification from a short abbreviation. Ask the responsible professional to resolve a clinical definition when necessary.
Freeze the data dictionary used for the comparison
Save the original export and its field definitions. Record which software version or form produced it and the dates when a new rule took effect. A historical blank field should not automatically acquire the meaning of a new mandatory checkbox.
| Question | Evidence to retain |
|---|---|
| Which animals were eligible? | Named cohort and inclusion rule |
| Which events counted? | Definition and original event entries |
| Which date selected the record? | Event, entry or export date |
| What was missing? | Missing count for each required field |
| Did the rule change? | Earlier and later dictionary editions |
The STROBE reporting checklist, items 7–9 and 12, calls for clear variables, assessment methods, attention to bias and an explanation of missing-data handling. These are general reporting principles; the checklist does not supply a farm benchmark or certify a reproductive interpretation.
See how missing litters alter an average
Consider an entirely fictional export of ten eligible litter records. Eight contain the selected count and their values sum to eighty. Two are blank. The average among the eight completed records is 80 ÷ 8 = 10. Dividing the same sum by all ten gives 8, but that calculation has silently assigned zero to both missing entries.
Report the completed-record average with its coverage: eight of ten eligible records. Keep the two missing entries visible. Neither average establishes the true value for the full ten-record cohort. If the missing records are later recovered, issue a dated revision with the recovered values rather than erasing the earlier limitation.
Check who disappears between stages
Reconcile the number eligible, the number with complete entries and the number analyzed. List exclusions with reasons. Do not remove an animal from the original cohort simply because it lacks the outcome that would make the final graph easy to calculate.
A transferred animal may have a known departure but an unknown later outcome. Those are separate facts. Likewise, a duplicate record correction should change the analysis once, with a traceable explanation. It should not look like a new reproductive event on the correction date.
Compare like groups before telling a performance story
Inspect the group composition behind an overall average. In another fictional example, an earlier export contains mostly first-parity animals and a later export mostly other parities. A difference between the combined values does not isolate a change within either category. Present the actual composition and compatible subgroup summaries before attributing the change to a management decision.
Keep worker coverage, retrospective entries and changed definitions in the limitations. This review can identify a reporting problem; it cannot determine why a reproductive outcome occurred. Bring clinical questions to the veterinarian with the original evidence.
Use the pig reference hub and incidence and prevalence guide for related reading. A useful comparison makes its selection and arithmetic recoverable, even when the resulting conclusion is that the available records cannot support a ranking.
An unchanged abortion count does not resolve a litter-size decline
Assume two fictional sets each contain twenty complete litter records under an unchanged selected birth-count definition. Their totals are 240 and 220, giving means of 12 and 11 per recorded litter. A separate calendar-period abortion register has two entries in each period. Those equal counts do not explain the lower litter mean or establish an unchanged abortion rate; the eligible pregnancy populations and event definitions must also be known.
Keep the farrowing cohort, register categories, parity composition and actual date windows beside both summaries. Do not substitute the calendar register for the litter cohort or treat a stable field as proof that another reproductive concern is absent. Give the veterinarian the supported figures and unresolved definitions rather than assigning a mechanism from the combination of two dashboard trends.
Cite this reference
LivestockDefense. “Check the recording rules behind sow reproductive figures.” Updated 2026-10-31. https://livestockdefense.com/guides/recognizing-recording-bias-in-sow-reproductive-kpis/
Add the date you accessed the page. This identifies the current reference; keep a permitted saved copy if your work depends on its exact wording. A citation alone does not establish support for a claim.
How to cite a page and its revision ↗Sources & boundaries
Read the evidence in context.
These documents support the statements identified in this reference. Check their scope and dates when deciding whether they answer your question.
- STROBE Initiative: STROBE Statement: combined observational-study reporting checklist
Source update: Not provided. Most recent check: 2026-10-30.
Items 7–9 and 12–14: variable definitions, measurement, bias, missing-data handling and participant flow. General reporting concepts only; no livestock benchmark, diagnostic definition or treatment effectiveness verified.
